Disney Builds Three Star Wars Robots That Choose Their Own Lines
The Anzellans watch guests in real time and pick what to say and how to move, with no puppeteer and every line written by Imagineers.

Three tiny Star Wars mechanics named Vilo Vele, Kado Bre'a and Jibu Bre'a can see what you are doing and decide, on the spot, what to say back. Walt Disney Imagineering revealed them on November 18 as its newest Audio-Animatronics prototype, built with Amazon Web Services and run without a puppeteer. For a craft that has played the same choreographed show for six decades, it is the first glimpse of park characters that answer the guest in front of them.
No Fixed Loop, No Puppeteer
Audio-Animatronics have always been performers with perfect memory. Disney's own description is blunt: traditional figures have performed carefully choreographed shows, with the same performance, in the same order, every time. The Anzellans break that rule. They do not run on a fixed loop, Disney says, but decide for themselves what to say and how to move.
Each character perceives what guests are doing in real time and reacts to them, and also to the other two Anzellans sharing its workbench. There is no one off stage steering them with a joystick. The trio looks the part: small, bearded and goggled, the same alien species as Babu Frik, the pint-sized droid fixer from The Rise of Skywalker.
“At Disney, we always start with the story and how we want guests to feel. With the Anzellans, that feeling is surprise.”— Kyle Laughlin, Walt Disney Imagineering
Surprise is easy to promise and hard to engineer. So what is actually doing the deciding?
Fast Models for Reflexes, Larger Ones for the Scene

The intelligence behind each character runs on AWS. Disney describes a split that mirrors how people react: fast-decision models handle the quick reactions, while larger models weigh the wider picture of what is happening around the workbench. Every component is tuned for latency, because a character that answers two seconds late stops feeling alive.
What the AI does not do is invent dialogue. The system picks from pre-recorded audio written by Disney's Imagineers and plays it with animations generated to fit the moment. The choice of line and the movement are live; the words themselves are authored.
Samira Panah Bakhtiar, who leads media and entertainment at AWS, put it plainly: every word, every gesture and every personality trait was crafted by Disney's Imagineers.
That division keeps the characters inside Star Wars and inside Disney's voice. It also leaves the question of when guests outside a tech conference get to meet them.
First Stop: A Showroom in Las Vegas
For now the Anzellans are a research prototype from Walt Disney Imagineering R&D, and their first public appearance is at AWS re:Invent in Las Vegas, where attendees can meet them at the Customer Showcase at Caesars Forum from November 30 to December 4. Disney calls them characters guests can one day meet, and has not named a park or a date.
AWS calls the project an example of physical AI: models that sense a room and act in it, not only in a chat window. For theme parks, the payoff is a character that notices a child waving and waves back on purpose. The Anzellans are the first Disney figures built to stop performing at guests and start playing with them.
From Fixed Show to Live Reply
The Brief
Stay curious
AI and technology: what changes and why it matters.
Your daily selection, in English or Spanish.
Free forever. Unsubscribe anytime.
More stories

The Specialty News





Conversation
Start the conversation